LTR Framework - How to Train an Army of Models for Efficient Ranking Experiments

Discover how OLX tackled the challenge of experimenting with rankings across multiple countries and categories in this 45-minute conference talk from Haystack EU 2024. Learn about the implementation of the LTR Framework, which enabled Data Scientists to adjust configurations, select features, and optimize targets efficiently. Explore the Model Store concept for deploying models in A/B ranking tests, and understand how this solution addressed bottlenecks in experimenting with numerous models. Gain insights into the benefits of rapid testing, including accelerated growth of successful events and improved understanding of user relevance preferences. Presented by Marcin Gumkowski, Senior Machine Learning Engineer, and Catarina Gonçalves, Junior Data Scientist at OLX Group, this talk offers valuable knowledge on building and improving Search products using Machine Learning and Cloud Computing, as well as solving ranking problems in large-scale environments.

Haystack EU 2024 -Marcin Gumkowski & Catarina Gonçalves:LTR Framework-how to train an army of models